RESEARCH

GEOCHEMISTRY on the basis of x-ray diffraction data that are presented here (Fig. 1). -VII is a high-pressure form of ice that is stable above 2.4 GPa Ice-VII inclusions in : (18). As we show, ice-VII (along with magnesian calcite, ilmenite, and halite) sensitively records high remnant pressures, which then also constrain Evidence for aqueous fluid the pressure and temperature where it has been encapsulated in the host crystal, similar in Earth’s deep mantle to other micrometer-scale inclusions of soft molec- ular materials such as CO2,CO2-H2O, and N2 in diamond (19–23). O. Tschauner,1* S. Huang,1 E. Greenberg,2 V. B. Prakapenka,2 C. Ma,3 G. R. Rossman,3 By retaining high pressures, ice-VII inclusions A. H. Shen,4 D. Zhang,2,5 M. Newville,2 A. Lanzirotti,2 K. Tait6 monitor the former presence of H2O-rich fluid at different depths in the diamond-bearing man- Water-rich regions in Earth’s deeper mantle are suspected to play a key role in the global tle. Remnants of former fluids and melts have water budget and the mobility of heat-generating elements. We show that ice-VII occurs been found as inclusions in many diamonds as inclusions in natural diamond and serves as an indicator for such water-rich regions. through infrared (IR) spectroscopy and micro- Ice-VII, the residue of aqueous fluid present during growth of diamond, crystallizes upon chemical analysis (19, 22). On the basis of IR ascent of the host diamonds but remains at pressures as high as 24 gigapascals; it is spectroscopy, a lower-pressure ice phase, VI, has now recognized as a by the International Mineralogical Association. In particular, been reported as an inclusion in diamond (24), ice-VII in diamonds points toward fluid-rich locations in the upper transition zone and but the fact that aqueous fluid has been trapped around the 660-kilometer boundary. Downloaded from in the TZ or LM and crystallized as ice-VII was previously unknown. he water content of Earth’smantleisa U.Previousmodelshaveproposedthatupwell- Diamonds from southern Africa (Orapa, Nama- key parameter of Earth’s water budget (1). ing mantle releases water while crossing the TZ- qualand), China (Shandong), Zaire, and Sierra Global recycling of water in Earth drives UM boundary and generates a buoyant layer of Leone were examined by diffraction with hard important forms of volcanism such as is- a comparatively water-rich melt that accounts x-rays (0.3344 Å) and a beam focused to 2 mm×3mm

T ’ http://science.sciencemag.org/ land arcs (2, 3), controls upper-mantle rhe- for much of Earth s budget of heat-generating at the undulator beamline 13-IDD (GSECARS, ology (4), and plays a role in the evolution of elements (11). Further, downward-moving mate- Advanced Photon Source, Argonne National Lab- mantle plumes (5). Subducted oceanic crust de- rial that crosses the TZ-LM boundary may lose oratory). A PILATUS 3X CdTe1M pixel array de- hydrates at shallow depth (2, 3), whereas faster its chemically bound water upon transformation tector was used for collecting diffraction data. and colder slabs can carry water much deeper into the LM bridgmanite and periclase, Diffraction from ice-VII was observed in dia- (6). Average water abundance in the mantle thus triggering mantle metasomatism around mondsfromsouthernAfricaandChinaasisolated changes markedly across the boundary between the 660-km boundary (10, 11). arrays within the diamond matrix of dimensions the upper mantle (UM) and the transition zone The actual water content of different mantle ranging from 3 mm×10mmtolessthan2mm× (TZ),aswellasacrosstheboundarybetweenthe regions depends not only on thermodynamic lim- 3 mm. Ice-VII was observed as isolated inclusions TZ and the lower mantle (LM). These boundaries its on water solubility but also on the mechanism tens of micrometers afar from other inclusions, originate in the pressure-driven phase trans- in real Earth that carries water to beyond 410 km in proximity to small amounts of nickeliferous

formations of mantle rock minerals—in partic- depth (1). The question of water abundance in carbonaceous iron (Fig. 1), similar to metal in- on March 8, 2018 ular, the transformations of olivine to wadsleyite thedeepermantlecanthereforeonlybedecided clusions reported in earlier studies on diamond at 410 km and from ringwoodite to bridgmanite on the basis of actual samples from these re- (24),toilmenite,and,inonecase,toalkalihalides. and periclase at 660 km depth. The UM contains gions whose mineralogical or petrographic record Many inclusions, such as silicates, carbonates, only modest amounts of water on average (7), is placed in context with geochemical and geo- oxides, and halides, were observed through x-ray whereas the average water content in the TZ has physical observations, such as seismic wave at- fluorescence mapping, some of which could be been estimated at ~0.1 weight percent (8), or more tenuation and electrical conductivity (1, 4, 8), clearly identified through diffraction or micro- than 10 times that of the UM. The abundance of and by using experimentally obtained thermo- chemical analysis (Table 1). The observation of water in the LM is unknown, but its constituent dynamic properties. Diamond is the main source olivine with 94 to 97 mol % forsterite compo- minerals appear to have much lower solubility of of minerals from the deeper mantle: A mineral- nent (Fo 94-97) places the origin of the hosting water than the minerals of the TZ (9, 10). The ogical record from depths as great as 660 km or diamonds in mantle peridotite rather than ec- average abundance of chemically bound water even beyond has been identified as inclusions logite, whereas ilmenite is indicative of meta- in these different regions of Earth, as well as pos- in natural diamonds (12, 13). Pearson et al.(14) somatized mantle (15). sible occurrences of smaller layers or loci of water- reported an inclusion of hydrous ringwoodite in In all cases, the diffraction pattern of ice-VII rich rock or melt, are central to our understanding diamondfrombelow520kmdepthwhosefor- is powderlike with no visible granularity (Fig. 1, of Earth’s water budget over extended geologic mation implies a much more hydrous environ- inset). The patterns were unambiguously iden- time. Furthermore, fluids and water-assisted ment than the UM average. Generally, peridotitic tified by Rietveld refinement as those of ice-VII partial melting mobilize mantle-incompatible diamonds mark regions of mantle metasoma- (Table 1 and Fig. 1) (26) and correlate with IR elements, including heat-generating K, Th, and tism (15). Over long geologic time, metasoma- absorption bands of O-H stretching vibrations tism is almost pervasive, at least in Earth’sUM. (26). The high quality of some of the diffraction 1Department of Geoscience, University of Nevada, Las Diamonds conserve important information about data resulted in noticeable proton contributions Vegas, NV 89154, USA. 2Center of Advanced Radiation these important processes (16). to the patterns (Fig. 1). Aqueous fluid in Earth’s Sources, University of Chicago, Chicago, IL 60637, USA. Here, we provide evidence for the presence of mantle is expected to be saline (22)andice-VII 3Division of Geology and Planetary Science, California Institute of Technology, Pasadena, CA 91125, USA. aqueous fluid in regions of the TZ and around can dissolve at least up to ~2 mole percent (mol %) 4Gemological Institute, China University of Geosciences, the TZ-LM boundary by showing the presence of alkali halides (27, 28). For structure analysis Wuhan 430074, China. 5School of Ocean and Earth ofice-VIIasinclusionsindiamondsfromthese and for the assessment of pressure, dissolution ’ Science and Technology, University of Hawai iatManoa, regions of the mantle. Ice-VII has recently been of (Na,K)Cl has been taken into consideration. Honolulu, HI 96822, USA. 6Royal Ontario Museum, Toronto, Ontario M5S 2C6, Canada. recognized as a mineral by the International We provide details about the crystallography of *Corresponding author. Email: [email protected] Mineralogical Association [2017-029 (17)] (Table 1) natural ice-VII in (26); here, we focus on the

Tschauner et al., Science 359, 1136–1139 (2018) 9 March 2018 1of4 RESEARCH | REPORT petrologic implications. We note that, with one These inclusions also contain high amounts of phases that have been encapsulated at the same exception (Table 1), the total salinity of the ob- silicate or carbonate (22). Many inclusions of sili- depth are precipitates from one complex fluid. served inclusions (ice-VII plus coexisting phases) cates, carbonates, oxides, and halides are found We propose that a complex aqueous fluid was is much lower than the alkali halide content of within distances of several tens to 100 mmfrom entrapped as separate inclusions that crystal- fluid inclusions commonly observed in diamonds. the ice-VII inclusions. Plausibly, some of these lized as ice-VII, carbonate, halide, and silicate rather than mimicking the bulk fluid composi- tion in each inclusion (26). Fig. 1. Diffraction pat- Because of their confinement by the rigid tern of ice-VII in dia- diamond host crystal, the inclusions of ice-VII mond M57666 from remain at high pressure, allowing us to use the Orapa. Black crosses equation of state of ice-VII to determine mini- are data points; the mum pressures for formation of the surround- Rietveld refinement ing diamond (Table 1). We found pressures of based on this pattern about6GPaand9±1.6GPafordiamondsfrom of the type material Orapa, Botswana. We determined a pressure of (blue curve) converged 12 ± 2 GPa for a diamond from Shandong, to a weighted profile China, and 24 to 25 (±3) GPa for a specimen refinement parameter from Namaqualand. In the fibrous rim of one

Rwp of 5.72% with a diamond, we found a NaCl hydrate at a pressure profile refinement of at least 1 GPa, rather than ice-VII (Table 1). parameter Rp of 4.57% We note that this sample was found at the same Downloaded from and with c2 = 1.71 for locality as three of the samples with inclusions 1398 observations. The around 6 and 9 GPa (Orapa, Botswana). Thus, structure factor–based ice-bearing diamonds from one location are not refinement parameter fromthesamesourceregionorunderwentad-

RF was 4.0% (see table ditional growth at shallow depth. However, dia- S3). The green line monds from geologically different locations such

indicates the residual as China and southern Africa contain ice inclu- http://science.sciencemag.org/ of fit. Blue tick marks sionsthatresideathighpressuresof9to12GPa denote symmetry-allowed reflections of ice-VII; olive tick marks denote allowed reflections of (Table 1). Differences in nitrogen aggregation (Fe,Ni,C). (Fe,Ni,C) contributes 0.75 ± 0.05 volume percent to the pattern. The equation of state (27) also indicate markedly different geologic ages allows estimation of the current pressure of this inclusion of ice-VII as 9.2 ± 1.6 GPa. The inset is a or temperatures of the host diamonds and overall diffraction image of ice-VII. A diffraction image of ice-free diamond ~20 mm afar was used as a low nitrogen aggregation (26). Of 13 occurren- background image. Remaining single-crystal reflections are from the hosting diamond. ces of ice-VII, eight fall into a narrow pressure

on March 8, 2018

Fig. 2. Pressures of occurrences of ice-VII in natural diamonds. the intersection of fluid H2O isochores and mantle adiabats. We (A) Current residual pressures (black squares; error bars denote SD). consider the current average mantle geotherm (25) and two adiabats These residual pressures also represent lower bounds of the that are 300 K hotter and cooler as a reference frame. The intersection pressures of entrapment. Thirteen inclusions have been identified with of the isochores with these adiabats gives ranges of pressure of pressures ranging from >1 to 25 (±3) GPa. The majority of inclusions encapsulation of the aqueous fluid in growing diamond. We also have residual pressures around 9 GPa. Red lines indicate the transitions show the entrapment path of ilmenite (black diamonds; error bars between UM and TZ and between TZ and LM. (B) Reconstruction of denote SD) with a residual pressure of 11 to 12 GPa that was found in entrapment conditions. We used the equations of state of diamond and ice-bearing diamond GRR1507 (Table 1). Simultaneous entrapment fluid H2O(26) to estimate plausible entrapment conditions. Current of ilmenite and fluid H2O occurred at a pressure-temperature regime of pressures were corrected for elastic relaxation of surrounding diamond 450 to 550 km depth and 1400 to 1900 K. The melting curve of ice-VII (33). Within uncertainties, entrapment conditions can be estimated by (26) is given for reference.

Tschauner et al., Science 359, 1136–1139 (2018) 9 March 2018 2of4 RESEARCH | REPORT

Table 1. Occurrences of ice-VII inclusions in diamond, number of occur- depth was taken as an average of the two measured dimensions. A volume rences, current pressure, and other observed inclusions. The assess- of 40 mm3 corresponds to occurrences in single patterns. For weak single 3 ment of pressure and its correction for elastic relaxation (Pcor) is described patterns, we assume a volume much smaller than 40 mm . Coexisting

in (26). (Na,K)Cl·5H2O has an ice-VII–like structure with Na + K + Cl residing phases are defined as those observed in the same patterns as ice-VII; the on sites 2a and 4b (26). Volumes (mm3) are approximate values based on phases may not be hosted by the same cavity. Errors in the last significant integrated diffraction signal over the two-dimensional sampling grid; the digit(s) are shown in parentheses.

O-H stretch Ice-VII (Na,K)Cl Volume Specimen P (GPa) P (GPa) and combination Coexisting phases Other phases volume (Å3) (mol %) cor (mm3) bands (cm−1) ≥ — GRR1518...... 88.0(1) 16.8 1 3220(20) <40 (Na,K)Cl, (Na,K)Cl·5H2O † ‡ ...... GRR1507 33.689(8) 3(1) 6(1) 7(2) 3210(20) <<40 Ilmenite 85% Olivine (Fo94-97), calcite, sellaite — † ...... GRR1507 32.40(1) <5 7.9(1.4) 9(2) <40 Ilmenite 81% — ...... GRR1521 31.647(5) <0.1 9.2(1.6) 11(2) 3000 to 3100 <40 Garnet, olivine

...... M57666* 31.653(4) 4.4(6) 9.2(1.6) 11(2) 2950 to 3100 3000 (Fe,Ni,C) 0.75% — ...... GRR1521 31.632(5) 2.4(6) 9.3(1.6) 11(2) 500 (Fe,Ni,C) 59.7% — ...... GRR1521 31.658(8) 1.8(1) 9.2(1.6) 11(2) 500 (Fe,Ni,C) 0.7% — ...... GRR1521 31.641(5) <0.1 9.3(1.6) 11(2) <40 (Fe,Ni,C) 1.0% — Downloaded from ...... GRR1521 31.671(5) 3.6(4) 9.2(1.6) 11(2) <40 (Fe,Ni,C) 1.3% — — ...... GRR1521 31.6705(9) <0.1 9.2(1.6) 11(2) 3000 — ‡ ...... Balas-1 26.38(4) 3(1) 24(3) 28(5) 3140(20), 4500(50) <40 Calcite, halite, sellaite, chromite — — ...... Balas-3 26.25(9) 3(2) 25(3) 28(5) <<40 — ...... SM548 30.284(9) 2 to 5 12.0(2.0) 14(2) 3100(50) <40 Calcite,§ halite,§ olivine

† a b c ‡ *Ice-VII type specimen. (Ilm92-97 Geik3-8), = = 5.0161(6) Å, = 13.686(4) Å. Sellaite, MgF2, olivine composition was measured by SEM-EDS from four http://science.sciencemag.org/ different locations. §(Ca0.75(2)Mg0.25(2))CO3, a = b = 4.803(2), c = 15.98(1) Å, which corresponds to 8 to 9 GPa (26); halite, a = 5.231 Å, which corresponds to 10(1) GPa (26).

interval around 8 to 12 GPa (Fig. 2A and Table 1). We argue that within given uncertainties of ther- The much lower compatibility of H2O in the This pronounced clustering of sustained pres- moelastic properties, the pressure range of en- LM relative to the TZ (9, 10) has been suggested sures and the observation of ice inclusions at trapment is constrained by the intersection of to cause mantle metasomatism when slabs or 24 to 25 GPa have geologic implications: They these adiabats, with pressure-temperature paths surrounding mantle are sinking to below the point toward regions in the deep mantle where of the inclusions approximated as isochores 660-km boundary: H2O that cannot be chem- aqueous fluid was present during growth of (19, 20)(Fig.2B);see(26) for details. We note ically bound by the bridgmanite-periclase phase

diamond. We note that the sustained pressures that our estimation of entrapment conditions for assembly of the lower mantle is released and on March 8, 2018 of ice-VII inclusions match those of inclusions ice-VII and ilmenite overlap in the depth of the interacts with surrounding mantle (10, 11). Our of ilmenite, magnesian calcite, and halite in the shallower TZ (Fig. 2B). observation of ice and its plausible origin from a same host diamonds (Table 1) (26), indicating Two main conclusions can be drawn: (i) The free aqueous fluid below 610 km depth (Fig. 2B) growth at similar depth. aqueousinclusionswereentrappedasfluidrath- is consistent with this hypothesis and connects We can attempt to go a step further and esti- er than ice (32). Crystallization into ice-VII the experimental and geodynamic work with ob- mate the pressure under which the inclusions has occurred at much shallower depth during servations from nature. More generally, natural were trapped by correcting for the thermal con- ascent. (ii) Despite marked uncertainties in the ice-VII, magnesian calcite, and halite provide new tribution to pressure and volume caused by the equations of state, the entrapment pressures for indicators for the presence of water-bearing and high temperatures in the deeper mantle. This the ice inclusions that are currently at ~8 to 12 GPa carbonaceous fluid in actual deep Earth samples, requires reconstruction of plausible paths that turn out to be sufficiently narrow to permit a which can be linked to geochemical and geo- connect the current residual pressure of these statementofthedepthoftheirsourceregions physical information obtained from the same re- inclusions with the pressures and temperatures (Fig. 2B): They range from 400 to 550 km depth. gions in Earth. Ice-VII and other micro-inclusions of their possible host regions in the deep Earth. For the inclusions at 24 to 25 GPa, the source (Table 1) provide, through their residual density, That temperature can be along the current region is less narrowly estimated at 610 to 800 km an accurate minimal pressure of formation, al- average mantle geotherm, a possibly higher depth; 620 km is the depth of entrapment esti- though the reconstruction of the entrapment geothermal temperature in the geologic past, or mated for dense N2 inclusions in diamond (20). pressure at mantle temperatures is currently lim- a cooler regime in the vicinity of subducted Overall, the ice-VII inclusions show directly that ited by the available thermoelastic data on aque- slabs or within harzburgitic restites. There- water-rich fluid occurs in regions within the TZ ous (saline) fluid. fore, we encompass these different possible and around the 660-km boundary, or possibly temperature-depth regimes by a reference frame in the shallow LM (Fig. 2B). Taken together with REFERENCES AND NOTES defined through three adiabats whose temper- theoccurrenceofmagnesiancalciteandhaliteat 1. S. Karato, in Treatise on Geophysics, G. Schubert, Ed. (Elsevier, – atures include possiblecoolandhotregimes similar pressures, we can infer the presence of 2015), pp. 105 144. 25 29–32 2. T. L. Grove, C. B. Till, E. Lev, N. Chatterjee, E. Médard, Nature ( , ): the modern average mantle geotherm complex aqueous, saline, and carbonaceous fluid 459, 694–697 (2009). (25) plus two adiabats 300 K below and above at those depths. At present, we cannot assess the 3. S. G. Nielsen, H. R. Marschall, Sci. Adv. 3, e1602402 (2017). this geotherm. This doesnotimplythattheice- extent of these fluid-rich regions, although it is 4. H. Jung, S. Karato, Science 293, 1460–1463 (2001). – bearing diamonds formed in a convecting mantle; plausible that they were constrained both in 5. N. Métrich et al., J. Petrol. 55, 377 393 (2014). 6. L. H. Rupke, J. P. Morgan, M. Hort, J. A. D. Connolly, Earth rather, it defines a temperature-pressure range space and time before evolving into less hydrous Planet. Sci. Lett. 223,17–34 (2004). that includes different plausible source regions. partial melts. 7. D. R. Bell, G. R. Rossman, Science 255, 1391–1397 (1992).

Tschauner et al., Science 359, 1136–1139 (2018) 9 March 2018 3of4 RESEARCH | REPORT

8. S. Karato, Earth Planet. Sci. Lett. 301, 413–423 (2011). 25. J. M. Brown, T. J. Shankland, Geophys. J. R. Astron. Soc. 66, contract DE-AC02-06CH11357. The Advanced Light Source is 9. T. Inoue, T. Wada, R. Sasaki, H. Yurimoto, Phys. Eartheart 579–596 (1981). supported through contract DE-AC02-05CH11231. Author Planet. Inter. 183, 245–251 (2010). 26. See supplementary materials. contributions: O.T. participated in design, interpretation, data 10. B. Schmandt, S. D. Jacobsen, T. W. Becker, Z. Liu, K. G. Dueker, 27. M. R. Frank, E. Aarestad, H. P. Scott, V. B. Prakapenka, collection, and analysis of the reported results, and in drafting Science 344, 1265–1268 (2014). Phys. Earth Planet. Inter. 215,12–20 (2013). and revising the manuscript; S.H. participated in interpretation of 11. D. Bercovici, S. Karato, Nature 425,39–44 (2003). 28. L. E. Bové et al., Proc.Natl.Acad.Sci.U.S.A.112, 8216–8220 (2015). the results and in drafting and revising the manuscript; E.G., 12. F. E. Brenker, T. Stachel, J. W. Harris, Earth Planet. Sci. Lett. 29. C. Herzberg, K. Condie, J. Korenaga, Earth Planet. Sci. Lett. V.B.P., C.M., G.R.R., A.H.S., D.Z., and K.T. participated in data 198,1–9 (2002). 292,79–88 (2010). collection and revising the manuscript; and M.N. and A.L. 13. T. Stachel, J. W. Harris, G. P. Brey, W. Joswig, Contrib. Mineral. 30. C. T. A. Lee, P. Luffi, T. Plank, H. Dalton, W. P. Leeman, participated in data collection. Competing interests: The – Petrol. 140,16 27 (2000). Earth Planet. Sci. Lett. 279,20–33 (2009). authors have no competing interests. Availability of data and – 14. D. G. Pearson et al., Nature 507, 221 224 (2014). 31. C. A. Dalton, C. H. Langmuir, A. Gale, Science 344,80–83 (2014). materials: Additional chemical and crystallographic information – 15. N. V. Sobolev et al., Lithos 39, 135 157 (1997). 32. C. R. Bina, A. Navrotsky, Nature 408, 844–847 (2000). about ice-VII is provided as supplementary material. Raw 16. S. B. Shirey et al., Rev. Mineral. Geochem. 75, 355–421 (2013). data are deposited at Dryad (doi:10.5061/dryad.7145v). 33. R. J. Angel, M. L. Mazzucchelli, M. Alvaro, P. Nimis, F. Nestola, 17. O. Tschauner, E. Greenberg, V. Prakapenka, C. Ma, K. Tait, Crystallographic and chemical information on type material Am. Mineral. 99,2146–2149 (2014). Mineral. Mag. 81, 1033–1038 (2017). ice-VII are deposited with the ICSD database. 18. V. F. Petrenko, R. W. Whitworth, Physics of Ice (Oxford Univ. ACKNOWLEDGMENTS Press, 1999), p. 253. SUPPLEMENTARY MATERIALS 19. O. Navon, Nature 353, 746–748 (1991). We thank four anonymous reviewers for their helpful comments, www.sciencemag.org/content/359/6380/1136/suppl/DC1 20. O. Navon et al., Earth Planet. Sci. Lett. 464, 237–247 (2017). and H. A. Bechtel for support at the Advanced Light Source. Materials and Methods 21. E. S. Izraeli, J. W. Harris, O. Navon, Earth Planet. Sci. Lett. 187, Funding: This work was supported by U.S. Department of Figs. S1 to S4 323–332 (2001). Energy (DOE) awards DESC0005278, DE-FG02-94ER14466, Tables S1 to S3 22. Y. Weiss, I. Kiflawi, O. Navon, in Proceedings of 10th International and DE-NA0001974, and by NSF grants EAR-1634415, References (34–59) Kimberlite Conference (Springer, 2013), pp. 271–280. EAR-1128799, EAR-1322082, EAR-0318518, and DMR-0080065. 23. H. Kagi et al ., Mineral. Mag. 64, 1089–1097 (2000). The Advanced Photon Source, a DOE Office of Science User 5 July 2017; accepted 19 January 2018 24. E. M. Smith et al., Science 354, 1403–1405 (2016). Facility, is operated by Argonne National Laboratory under 10.1126/science.aao3030 Downloaded from http://science.sciencemag.org/

on March 8, 2018

Tschauner et al., Science 359, 1136–1139 (2018) 9 March 2018 4of4 Ice-VII inclusions in diamonds: Evidence for aqueous fluid in Earth's deep mantle O. Tschauner, S. Huang, E. Greenberg, V. B. Prakapenka, C. Ma, G. R. Rossman, A. H. Shen, D. Zhang, M. Newville, A. Lanzirotti and K. Tait

Science 359 (6380), 1136-1139. DOI: 10.1126/science.aao3030

Encapsulating Earth's deep water filter Small inclusions in diamonds brought up from the mantle provide valuable clues to the mineralogy and chemistry of parts of Earth that we cannot otherwise sample. Tschauner et al. found inclusions of the high-pressure form of water called ice-VII in diamonds sourced from between 410 and 660 km depth, the part of the mantle known as the transition zone. The transition zone is a region where the stable minerals have high water storage capacity. The inclusions suggest Downloaded from that local aqueous pockets form at the transition zone boundary owing to the release of chemically bound water as rock cycles in and out of this region. Science, this issue p. 1136 http://science.sciencemag.org/

ARTICLE TOOLS http://science.sciencemag.org/content/359/6380/1136

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2018/03/07/359.6380.1136.DC1 MATERIALS

REFERENCES This article cites 54 articles, 16 of which you can access for free on March 8, 2018 http://science.sciencemag.org/content/359/6380/1136#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title Science is a registered trademark of AAAS.